480 likes | 534 Views
ARTIFICIAL IMMUNE SYSTEM. FERAT SAHIN & R.S. Srividhya KGCOE Colloquium Series II 01/12/2000. Outline. Are humans perfect machines? How the brain works? Human vs. Computer. Are computers designed with the correct logic!? Why biological based methods? Some history
E N D
ARTIFICIAL IMMUNE SYSTEM FERAT SAHIN & R.S. Srividhya KGCOE Colloquium Series II 01/12/2000
Outline • Are humans perfect machines? • How the brain works? Human vs. Computer. • Are computers designed with the correct logic!? • Why biological based methods? • Some history • Artificial Immune Systems KGCOE Colloquium Series II: Artificial Immune System
Are Humans Perfect Machines? • Learning • Continuous • Selective • Memory • Short term, Long term • Selective • Recall • Processing • Recognition (hearing, sight, various senses) KGCOE Colloquium Series II: Artificial Immune System
How the brain works? • One of the great mysteries of science: How the brain enables thought? • “Man has the largest brain in proportion to his size”, Aristotle. • The seat of consciousness, until the middle of 18th century. • The functional regions of the brain began to be mapped out, late 19th century. • Elements of the brain: • The neuron / nerve cell: the fundamental functional unit of all nervous system tissue, including the brain. • Each neuron consists of a cell body, or soma, that contains a cell nucleus. • There are number of fibers, called dendrites, and a single long fiber called axon branching out from the cell body. • The axon also branches into strands and substrands that connect to the dendrites and cell bodies of other neurons. • The connecting junction is called synapse. KGCOE Colloquium Series II: Artificial Immune System
How the brain works? • Signaling • Complicated electromagnetic reaction from neuron to neuron. • The synapses releases chemical transmitter substances • The chemical substances enter the dendrite, raising or lowering the electrical potential of the cell body. • When the potential reaches a threshold, an electric pulse or action potential is sent down to the axon. • Plasticity: long-term changes in the strength of connections in response to the pattern of stimulation. • Migration: sometimes entire collections of neurons can migrate from one place to another.* • Most of the information goes on in the cerebral cortex, the outer layer. KGCOE Colloquium Series II: Artificial Immune System
How the brain works? • Certain areas of the brain have specific functions • The third left frontal convolution of the cerebral cortex is important for speech and language - aphasia • The mapping between areas of the brain and the parts of the body they control , or from which they receive sensory input, • Radical changes of the mapping and multiple mappings. • How other areas can take over functions when one area is damaged is not fully known. Migration? No known intelligent system can perform this. • There is almost no theory about how an individual memory is stored. • Article about the face recognition of the brain. • Neurobiology is a long way from a complete theory of consciousness. • The only real alternative theory is mysticism: • “There is some mystical realm in which minds operate that is beyond physical science. THE SOUL!!!? KGCOE Colloquium Series II: Artificial Immune System
How the brain works? • A general comparison of the raw computational resources available to computers and brains. Computer Human Brain Computational units 1 CPU, 105 gates 1011 neurons Storage units 109 bits RAM, 1010 bit disk 1011 neurons, 1014 synapses Cycle time 10-9 sec 10-3 sec Bandwidth 109 bits/sec 1014 bits/sec Neuron updates 105 1014 • Even though the computer is a million faster in raw switching speed, the brains ends up being a million times faster at what it does. • Face recognition: • The brain requires less than a second - a few cycles. • A serial computer requires billions of cycles. KGCOE Colloquium Series II: Artificial Immune System
Why biological based methods? • Biological systems outperform the advanced machines • They are slower but effective • Face recognition • 4-5 cycles versus billions of cycles • 1 cycle of the brain is extremely slower than a cycle of a C • Storing a face requires Mega Bytes in a computer • Examples: • Genetic Algorithms • Neural Networks • Artificial Immune Systems KGCOE Colloquium Series II: Artificial Immune System
Learning: Computational and Biological Viewpoints • Computational viewpoint • Learning is about a method of representing functions using network of simple arithmetic elements, and about methods for learning such representations from examples • Biological viewpoint • The simple arithmetic computing elements correspond to neurons-the cells that perform information processing in the brain. • The network as a whole corresponds to a collection of interconnected neurons. • Besides computational properties, neural networks may offer the best chance of understanding many psychological phenomena that arise from the specific structure and operation of the brain. KGCOE Colloquium Series II: Artificial Immune System
Some History: The Foundations of AI • Philosophy (428 B.C.- present) • Mathematics (c. 800 - present) • Psychology (1879 - present) • Computer Engineering (1940 - present) • Linguistic (1957 - present)
Some History: AI from past to now • The gestation of artificial intelligence (1943 - 1956) • knowledge of the basic psychology and function of neurons in the brain • Turing’s theory of Computation. • Early Enthusiasm, great expectations (1952-1969) • Lisp • Adalines by Bernie Widrow, 1960 ( Enhanced version of Hebb’s learning) • Perceptrons by Frank Rosenblatt, 1960 (Perceptron Convergence Theorem) • A dose of reality ( 1966-1974) • Principle versus practice • Machine evolution (now called Genetic Algorithm) • Very large computational time, Combinatorial Explosion • Some fundamental limitations KGCOE Colloquium Series II: Artificial Immune System
Some History: AI from past to now • Knowledge-based systems: The key to the power? (1969-1979) • Expert systems: Medical diagnosis • Frames (Minsky, 1975): Collecting together facts about particular object and event types, and arranging the types into a large taxonomic hierarchy analogous to a biological taxonomy. • AI becomes an industry (1980-1988) • R1: the first commercial expert system. • The Fifth Generation project, by Japanese, to build intelligent computers • The Microelectronics and Computer Technology Corporation • Chip design and human-interface research • The booming AI industry: • Software: Carnegie Group, Inference, Intellicorp, and Teknowledge • Hardware: Lisp machines Inc., Texas Instruments, Symbolics, and Xerox • The industry vent from a few million in sales in 1980 to $2 billion in 1988 KGCOE Colloquium Series II: Artificial Immune System
Some History: AI from past to now • The return of neural networks (1986-present) • Large collection of neurons = large collection of atoms in Physics. • Hopfield (1982): statistical mechanics to analyze the storage and optimization properties of networks. • David Rumelhart and Geoff Hinton: the study of neural net models of memory. • Reinvention of Back-propagation algorithm ( mid 1980s) • AI versus neural networks: “AI Winter” • The fear • Historical factors KGCOE Colloquium Series II: Artificial Immune System
Some History: AI from past to now • Recent Events (1987-present) • Hidden Markov Models (HHMs): successfully applied in Speech • Judea Pearl’s (1980) Probabilistic Reasoning in Intelligent Systems • The belief networks formalism was invented to allow efficient reasoning about the combination of uncertain evidence. • They are claimed to be the best representation of the human belief and reaoning. • Normative expert systems by Judea Pearl , Eric Horvitz and David Heckerman: • “Ones that act rationally according to the laws of decision theory and do not try to imitate human experts.” - Think rationally and act rationally. • Distributed intelligent systems • Internet computing, mobile robots, autonomous and collaborative systems KGCOE Colloquium Series II: Artificial Immune System
Artificial Immune System • Introduction to human Immune system • The human immune system • Types of immunity • Type of immune system • Features of vertebrate immune system • Artificial immune system and properties • Application of immune system Artificial Immune System
Introduction to Immune systems The human immune system • is a natural defense mechanism • maintains the system against dynamically changing environments • sophisticated information processors • learns to recognize patterns • cells do the job of encoding, controlling the system in parallel • immune system is a distributed system with no central controller KGCOE Colloquium Series II: Artificial Immune System
The Human Immune System The main function of the human immune system • is to protect our body from infectious agents such as viruses, bacteria, fungi, and other parasites. The basic components of the immune system • are the lymphocytes or the white blood cells • Two types of lymphocytes B- lymphocytes T- lymphocytes KGCOE Colloquium Series II: Artificial Immune System
The Human Immune System B-lymphocytes • are produced by the bone marrows • roughly there are 10 million B-lymphocytes in the human body. KGCOE Colloquium Series II: Artificial Immune System
Distinct chemical structures and produces many Y shaped antibodies from its surfaces Ag Paratop Idiotop Epitop Ab The Human Immune System KGCOE Colloquium Series II: Artificial Immune System
Types of Immunity • Innate Immunity - Invertebrate immune system • it’s the natural resistance of the body to the foreign antigens. • Non-specific towards invaders into the body • Acquired Immunity - Vertebrate Immune system • Directed towards specific invaders • Immunological memory is modified by exposure to such foreign antigens. KGCOE Colloquium Series II: Artificial Immune System
Spreading influence of the antigen KGCOE Colloquium Series II: Artificial Immune System
Spreading influence of the antigen • When an antigen enters the body the B cells binds it • B cell analyzes the antigen and also creates new B cells • Each B cell passes the antigen onto other B cell objects within its neighborhood • The number of neighbors which are presented with the antigen depends on how many cells have already possessed the antigen • Antigen spreads through the network gradually decreasing in concentration as it goes KGCOE Colloquium Series II: Artificial Immune System
Types of Immune systems Two types of immune systems are 1. Vertebrate – lymphocytes Involves lymphocytes, which are antigen specific Different receptors for difference antigens 2. Invertebrate immune systems – phagocytes Involves Phagocytes, which is non-specific immune response No distinct receptors for specific antigens Tries to kill any antigen Malfunction of this system: Leukemia. B-cells attacks the blood cells as if they are foreign. KGCOE Colloquium Series II: Artificial Immune System
Features of vertebrate Immune system • Feature extraction – to determine the unique nature of the antigen. • Learn to recognize new patterns/antigens. • Work as distributed pattern recognizer. • Use content addressable memory to retrieve known patterns/antigens. Learning!!! • Use of specific proliferation and self-replication for quick recognition and response. Reproduction!!! • Eliminate/neutralize the effect of antigens in a systematic pattern. KGCOE Colloquium Series II: Artificial Immune System
Artificial Immune system Inspiration to engineering sciences • Performing complex tasks such as learning, memory of large number of components, immunity development over time. • Artificial Immune system is essentially for the imitation of the immune system properties to computers and application to various fields KGCOE Colloquium Series II: Artificial Immune System
Properties of Immune system applicable to AIS • Clonal selection Principle • Reinforcement learning • Immune memory • Jerne’s idiotropic network theory • Positive and negative selection • Affinity maturation • Self organization KGCOE Colloquium Series II: Artificial Immune System
Clonal Selection Principle Clonal selection principle • Only those cells that recognize the antigens reproduce • New cells are copies of their parents (clone) cells • Elimination of newly differentiated lymphocytes • Proliferation and differentiation on contact of mature cells with antigens KGCOE Colloquium Series II: Artificial Immune System
Jerne’s Idiotropic’s Network Jerne’s hypothesis states that • antibody does not exist independently in living organisms • communicate with each other through idiotope and paratope • The portion of the antigen and the antigen recognized by the antibody is called the epitope • The one on the antibody that recognizes the corresponding epitope is called paratope. • Antibodies also have antigenic characteristic called idiotope. KGCOE Colloquium Series II: Artificial Immune System
Stimulation Suppression Jerne’s Idiotropic’s Network Ag B Cell #2 Id2 B Cell #1 Ab2 P2 Id1 B Cell #3 Ab1 P1 Id3 Ab3 P3 KGCOE Colloquium Series II: Artificial Immune System
Immune memory Immune memory • Immune system remembers the already entered or attacked antigen • Primary response – system evokes the antibodies • Secondary response – remembers the attacked antigen • More rapid • shorter lag phase • higher rate • longer persistence of antibody synthesis KGCOE Colloquium Series II: Artificial Immune System
Immune memory Cross reactive response • Uses the property of associative memory • For two similar antigens , immune system responds faster to the second by associating the response with the first antigen • It is found useful in artificial intelligence and neural networks KGCOE Colloquium Series II: Artificial Immune System
Foreign proteins Fungi Viruses Bacteria Parasites Vertebrate host B-cell T-cell T helper (Th) cells T Cytotoxic (CTL) cells T-cell Lymphokines Antibodies Th2 Th1 T-cell Lymphokines HUMORAL IMMUNE RESPONSE Macrophages Interleukines CELL MEDIATED IMMUNE RESPONSE An overview of immune system KGCOE Colloquium Series II: Artificial Immune System
Antigen Pattern Recognition B11 B21 B31 T11 T21 T31 Clonal Selection Clonal Selection B31 B31 B31 T31 T31 T31 Cell-mediated Response Anti-bodies Humoral Response B31 B31 B31 T31 T31 T31 Memory T-Cells Memory B-Cells An overview of Immune system KGCOE Colloquium Series II: Artificial Immune System
Autonomous Multi-Agent Systems • Distributed Artificial Intelligence (DAI) As a sub field of AI, it has existed for less than two decades. DAI is concerned with systems that consists of multiple independent entities that interact in a domain. Two sub disciplines of DAI • Distributed Problem Solving (DPS), • Multi-Agent Systems (MAS). KGCOE Colloquium Series II: Artificial Immune System
Autonomous Multi-Agent Systems • Deals with behavior management in collections of several independent entities, or agents. • There are many definitions for an Agent: • An agent is an entity with goals, actions, and domain knowledge all situated in an environment • The way the agent acts is its behavior, and there should be an interaction between this behavior and the environment that surrounds him. KGCOE Colloquium Series II: Artificial Immune System
Dog and Sheep problem For Simplicity Sheep Dog Dog task: Force the sheep to return to the pen Sheep task: Avoid the dog KGCOE Colloquium Series II: Artificial Immune System
Dog and Sheep problem Distance ( Sheep, Pen ) Distance ( Dog , Sheep) Distance ( Dog , Sheep) Distance ( Dog , Pen ) Sheep Dog Adaptation Adaptation Direction Direction KGCOE Colloquium Series II: Artificial Immune System
Dog and Sheep simulation Two Dogs & one sheep S D2 D1 Area of focus KGCOE Colloquium Series II: Artificial Immune System
Dog and Sheep Identical behavior KGCOE Colloquium Series II: Artificial Immune System
Dog and sheepD2 solely different behavior KGCOE Colloquium Series II: Artificial Immune System
Board Game and AIS Antigen B cells KGCOE Colloquium Series II: Artificial Immune System
Board Game and AIS Antigen B cells KGCOE Colloquium Series II: Artificial Immune System
Board Game and AIS KGCOE Colloquium Series II: Artificial Immune System
Negative selection,self/non self learning Immune network dynamics/negative selection principle Negative selection principle Distribution, self organization Anomaly detection-computer security in terms of viruses, unauthorized user detection and elimination Image inspection &image segmentation Novelty detection algorithm for time series data to exhibit the normal behavior of the system Agent based approach,Network intrusion detection Application of AIS KGCOE Colloquium Series II: Artificial Immune System
Immune System Property Dynamic decentralized consensus making mechanism, Jernes network, clonal selection algorithm and network dynamics Petri net concepts Immune diversity, network theory and clonal selection principle Cross reactive memory, recruitment mechanism Application to artificial immune system Robots, Mutual interaction between modules,interaction between robot environments, autonomous agents Adaptive control, identification and synthesis, sequential control Optimization Neural network approaches Application of AIS KGCOE Colloquium Series II: Artificial Immune System
Genetic mechanisms, clonal selection principle, affinity maturation, content addressable memory, matching mechanisms, and network self organizing properties. Immune networks Pattern recognition-classification, prediction, diagnosis and data mining Sensor based diagnosis Application of AIS KGCOE Colloquium Series II: Artificial Immune System
Thank you KGCOE Colloquium Series II: Artificial Immune System